Modification of scale effect on constitutive model parameters using super large-scale triaxial tests and in-situ monitoring data

The significant discrepancy between the deformation values calculated based on the traditional large-scale triaxial test results on rockfill materials (with a maximum particle size of 60mm) and the monitoring data from the prototype dam raises concerns about the factors causing this difference and their respective influences. This paper investigates the effects of scale effect (including scaling methods and particle size effects) on the mechanical properties of rockfill materials in five world-class high dams using the first super large-scale triaxial apparatus in China (with maximum particle sizes of 200mm and 160mm) as well as a large-scale triaxial apparatus. The study also examines the influence of constitutive models (such as the Duncan E-B and E-μ models, and generalized plasticity model) on the calculation of dam deformations. Additionally, by incorporating monitoring data from the Lianghekou and Aertashi dam, the paper explores the regularities of scale effects and the rationality of constitutive models, thereby clarifying that errors related to scaling methods, particle size effects, and constitutive models are the primary causes of distortion in the calculations of high rockfill dams. The research findings indicate that under the same void ratio, the mixed-scale method tends to overestimate the modulus of rockfill materials compared to the similar grading method. Moreover, the traditional large-scale triaxial tests based on the similar grading method tend to overestimate the modulus of rockfill materials compared to the results obtained from the super large-scale triaxial tests, which effectively eliminate the size effect. Furthermore, the calculations based on the generalized plasticity model, particularly in terms of horizontal displacement, are more in line with actual observations than those based on the Duncan E-B and E-μ models. Finally, a scale effect classification correction method for model parameters is proposed specifically for the Duncan E-B and E-μ models. The research outcomes contribute to a deeper understanding of the scale effects of rockfill materials and the influences of different constitutive models. They also provide reliable experimental and numerical calculation references for deformation prediction in similar engineering projects.